Statistical Learning of Pediatric Mental Health-Related Emergency Department Visits Across COVID-19 Pandemic Periods

This article presents a statistical learning framework for studying the evolution of pediatric mental health-related emergency department (MHED) visit patterns across the pre-, during-, and post-COVID-19 pandemic periods using population-based administrative health records. The MHED records are formulated as zero-truncated recurrent event data, partitioned into three successive time periods. We develop the modeling framework in a stepwise manner, guided by model fit using a collection of MHED records. The resulting framework progresses from nonparametric marginal rate models to more structured Cox-type regression models for characterizing visit patterns. We ultimately apply stratified regression analysis to investigate changes in visit frequencies and covariate effects across pandemic periods, accounting for prespecified period cut-off points and coarsened individual follow-up information. The proposed framework is motivated by and illustrated using pediatric MHED data throughout the article, providing a practical approach for analyzing recurrent healthcare utilization data with evolving temporal patterns.